Deliverable 1.3 Horizontal Benchmarks – Analytics and Processing
This document is focusing on benchmarks in the horizontal layers according to the BDVA reference model with data visualization (visual analytics), data analytics and data processing.
Visual analytics is an area that has been less covered in existing benchmarks, but an existing starting point for this can be found in the Hobbit-IV benchmark on visualization and services, which also focuses on question answering and faceted browsing. Data analytics include a level of industrial analytics with descriptive, diagnostic, predictive and prescriptive analytics and the support for this with the use of machine learning. Machine Learning includes supervised and unsupervised learning as well as reinforcement learning and has a strong focus in many ongoing ICT14 and ICT15 projects. Analytics is addressed for graph representations in the Hobbit-II benchmark on Graphalytics, but is also a focus in benchmarks on deep learning like DeepMark and DeepBench. Different analytic benchmarks will typically address different big data types such as time series, spatial, image and text. The area of data processing architectures includes benchmarks for real time processing with stream processing, batch processing and interactive processing and main memory architectures. These are areas covered in many benchmarks such as BigBench, BigDataBench and SparkBench – benchmarking different processing architectures such as MapReduce (Hadoop), SPARK and Flink and others.
In this document, the benchmarks will be classified in the following categories:
- data visualization (visual analytics),
- data analytics – (including Machine Learning and AI benchmarks) – structured according to the Big data types
- data processing